The Daily WTF: Curious Perversions in Information Technology

DaveK · 2012-05-17 Reply Admin

ParkinT:
As I learned long, long ago (in what seems like a Galaxy far, far away) during my [first] career;
When troubleshooting a communications system walk the ISO model.

Or, if you know about algorithmic optimisation, binary chop it.

2012-05-17 Reply Admin

QJo:
James:
Sting? Are you DAFT? Hope I don't have to submit a DailyWTF based on your code! I would just love to debug code that performed dollar amount calculations in strings, that would be just lovely!
Of course they teach numerical analysis in American College Courses! But obviously, you failed "English"...

You may laugh, but when I was tasked with writing an app to extract numbers from an Oracle database and use a Java tool to squirt the results through a C socket into a FORTRAN program running on a VAX, the only straightforwardly reliable technique for passing a large number of floats back and forward was to pipe them across as strings. Relying on the internal format to be sufficiently consistent to allow us to write a converter (there wasn't a pre-built one available at the time) was too risky, and the effort in writing it wasn't worth the candle. Using the toString method at the Java end and FORMAT (F15.5) or whatever at the FORTRAN end was perfectly adequate.

yeah but converting floats to strings and visa versa isn't the same as using strings to perform float calculations which was the point ... this is no laughing matter

2012-05-17 Reply Admin

lucidfox:
TRWTF is floating point financial calculations.

Oh, really? No one's ever thought of THAT before!

You're that one chick, aren't you? Fuck, has any woman ever had an original though in her life?!

No wonder women are the new face of poverty - they just stand around and mimic what everyone else in the room ("everyone else" probably being all men, just to be realistic - unless it's a fucking fabric store) is saying and then bitching when it's not a huge hit.

I mean, who wants to have a beer and watch the game with that?

You want to be helpful, honey? Learn to suppress your gag reflex like a good whore. I'm so sorry, I wish I could stop, what's wrong with me!?!?!?!

2012-05-17 Reply Admin

Partly correct. There are leased lines that are dedicated to the end points. They are't copper all the way but the path between points A & Z are dedicated to those points. You can put your finger on this path. These were typically for voice grade data but there were a number of ringdown lines. The endpoints deal with the signaling. These are rare now days.

The most common leased lines were "virtual". There is a dedicated path between the end points and the switches. But once the call got to the switch it routed via shared trunking. These still exist in great numbers.

2012-05-17 Reply Admin

JichaelMackson:
16KB of RAM? Four times that, and it'll be all that I ever need.

Is this a DailyWTF meme? Because I saw a post in another thread talking about how "64K should be enough for everybody." The (supposedly, if it ever happened at all) original quote is about 640K, not 64K.

Maurits · 2012-05-17 Reply Admin

JJ:
JichaelMackson:
16KB of RAM? Four times that, and it'll be all that I ever need.
Is this a DailyWTF meme? Because I saw a post in another thread talking about how "64K should be enough for everybody." The (supposedly, if it ever happened at all) original quote is about 640K, not 64K.

Well, the difference between "64K" and "640K" is 0, which makes very little difference.

I must admit I don't understand why 640K would be a limit. 64K, sure (16-bit pointers, for example.) But 640K isn't even a power of 2.

EDIT: here it is: http://en.wikipedia.org/wiki/Conventional_memory#640_KB_barrier

2012-05-17 Reply Admin

QJo:
Norman Diamond:
QJo:
Using the toString method at the Java end and FORMAT (F15.5) or whatever at the FORTRAN end was perfectly adequate.
F15.5? Not E21.15? I hope my life doesn't depend on the accuracy of your conversions.
No, don't worry your fluffy little head about that, your life would *not* depend on such accuracy.
[...]

And no, it probably wasn't F15.5, I pulled the numbers out of the air without thinking about them too much, for illustrative purposes. It would probably be F16.3, which would have been sufficient to cover the numbers we were transferring.

F16.3 yields 3 decimal places of accuracy, which become all zeros for numbers smaller than 0.001 or which become stars for numbers greater than 1 trillion. That comes even closer to a spec which I was supposed to conform to before a customer was persuaded to change it. The resulting loss of accuracy would cause a collision avoidance system to produce meaningless results. It's a relief to know that lives did not depend on your system. It was also a relief when that customer was persuaded to allow more accuracy.

QJo:
But the fact that you suggested E21.15 suggests that you don't actually have much of a clue about the general problem domain - or that you're trolling.

Back at you. E21.15 provides 15 digits of precision regardless of what the exponent is. I hope my life will never depend on anything you produce.

2012-05-18 Reply Admin

Maurits:
I must admit I don't understand why 640K would be a limit. 64K, sure (16-bit pointers, for example.) But 640K isn't even a power of 2.

Wow, you actually had to look it up? Now I feel old....

2012-05-19 Reply Admin

This reminds where was telecommunications in our country during 80s. Wait time for landline 5-10 years. Manual connection even to 12km distances with extra paid. Even in 1991. No digital trunks,no fiber optics, data rejected by communist goverment only teletype and fax. Last analog phone exchange was disconnected around 2000. Local calls are still charged per minute and there never existed flat rate dialup.
We finally got landline in 1993, no tone dial usable, no tone dial on new Siemens EWSD. ADSL availibility started in 2004 in my area. I HATE YOU Communists and Telefonica O2. I missed adventures like this one because of them.

2012-05-20 Reply Admin

TGV:
flop:
I would have expected something like the 500 mile email ...
But well, we take what we get, right?
That's a great one!

Could have been worse. When I saw "statistics", I thought they might have just taken some sort of average of all the destinations that worked vs all the destinations that didn't.

2012-05-21 Reply Admin

F15.5! No ES21.15! No F16.3! etc...

Giggle. Youngsters of today...

Both F16.3 and E21.15 are very broken in one particular way for input. Consider what happens if (for whatever reason) there's no decimal separator in the input field.

(Note also that for input: F, E, D, F, ES, EN and G all do exactly the same thing, so if the input field does have a decimal separator, then you guys are just quibbling about the field width. Concerns about loss of precision when calling java's toString are certainly valid, but if you are trying to feed Fortran data edit descriptors to java's toString then you have more serious issues to deal with... )

2012-05-22 Reply Admin

PunchedCards:
F15.5! No ES21.15! No F16.3! etc...
Both F16.3 and E21.15 are very broken in one particular way for input.

Yes, that's why I learned to "roll my own" for input. And if a user could type input interactively, I even had to "roll my own" for integers, so I could discard invalid input and let the user try again, instead of crashing when the library rejected the input.

PunchedCards:
Concerns about loss of precision when calling java's toString are certainly valid, but if you are trying to feed Fortran data edit descriptors to java's toString then you have more serious issues to deal with... )

I suppose in my case the maker of the collision avoidance system might have been passing format strings in C syntax instead of Fortran syntax to the library that they were using, but I still had to persuade them to use a format with more than 3 digits of precision. When the spec called for me to conform to their 3 digits, the expected results were going to be pretty much useless.

2012-05-22 Reply Admin

("instead of crashing when the library rejected the input" -- that was in Fortran II, which didn't have ERR=exceptioncatcher)

2012-05-24 Reply Admin

Zylon:
ExChickenFarmer:
I'm the OP. Actually the computers were 8 Bit CP/M machines. They were custom built for our company (a WTF in itself). There were only 4 per plant so they just shared a dial-up phone line. It was very slow because with every screen it waited for the line to be free, dialed the mainframe, connected, submitted the request, received the data, and hung up.
Cripes. Yet another case of Remy anonymizing stories almost to the point of being pure fiction.

+1

And he even manages to make the stories less entertaining. He's levelling us with his meta-wtf's. give the man some respect.

Long Distance

Leave a comment on “Long Distance”